Errors in CGAP xProfiler and cDNA DGED: the importance of library parsing and gene selection algorithms

BMC Bioinformatics

Table 4 Change in probability values "P" reported by cDNA DGED and our algorithm when the display cut-off value "F" is changed are exemplified for three genes that are presented in the gene list when normal bone libraries are compared with cancerous bone libraries

UniGene Cluster ID	Name	Symbol	CGAP "P" values^a		Our "P" values^b
			"F" = 2	"F" = 3	"F" = 2	"F" = 3
164226	Thrombospondin 1, mRNA	THBS1	0.001	0.049	0.978	0.978
369397	CDNA FLJ53400 complete cds	TGFBI	0.001	0.045	0.982	0.982
462998	In-IGFBP-4 mRNA	IGFBP4	0.000	0.007	0.999	0.999

^a Calculated using on-line tools from CGAP; calculations based on equations (1 - 3) in this report). The calculated "P" value is close to zero (on a scale of zero to one) if the probability is high that the observed expression difference is genuinely greater than the user-specified "F" value, and is not due to sampling error [31, 33].
^b Calculated using equation (4) in this report. This produces a "P" value of between zero and one, but unlike the CGAP value, this is a decimal fraction of the likelihood of there being at least a threefold difference in expression of the transcript in the activated cells.

ISSN: 1471-2105